Hi,
On Tue, Nov 5, 2019 at 9:15 PM Thad Guidry <thadguidry(a)gmail.com> wrote:
Hi Team,
I often see missed results in the Quick Search input box when phrases
sometimes are indexed with an apostrophe, but its missed because we
sometimes just type simple titles or simple names like so:
"no mans sky"
(no results)
but when I click "containing... no mans sky"
then the results page displays
<https://www.wikidata.org/w/index.php?search=no%20mans%20sky&title=Special%3ASearch&fulltext=1&ns0=1&ns8=1&ns14=1&ns120=1&ns122=1&ns146=1&ns640=1&ns828=1&ns1198=1>
and shows the video game correctly.
It seems this might be because of an indexing choice for the Quick Search
input box?
(I know that Lucene-Solr can handle this for English input preference
https://github.com/apache/lucene-solr/blob/master/lucene/analysis/common/sr…
)
The reason you find "no man's sky" when searching "no mans sky"
using
Special:Search is simply because "mans" is matched somewhere in the
document (could be because of the stemmer or for example the Quora topic id
No-*Mans*-Sky-5).
The ApostropheFilter that you mention unfortunately has been designed for
the turkish language and would remove everything after the apostrophe which
is not what we want in general.
Completion (quick search input box) and Special:Search use very different
techniques, what we do for apostrophe in completion is transforming it into
a space.
Should we do something different for apostrophes in completion search? I
have no clue and this would require some evaluation, if you believe this is
important as it causes much annoyance feel free to file a task in
phabricator so that we can discuss about it.
David.